An Improved SIOU-YoloV5 Algorithm Applied to Human Flow Detection

At present, object detection technology in computer vision is widely used to monitor an area and obtain the information of the scene. This study is aimed at stampede accidents, using this technology to detect the size of the human flow. In this paper, based on the basic framework of YOLO, in order to obtain more detection accuracy and stability in the identification of human traffic, a variety of LOSS functions are improved to improve the detection performance of the same scene. Then, the results obtained after these improvements were evaluated, and compared with the original YOLOV5, it can be concluded that the performance of various improved methods in the human flow detection is better, the target detection is more accurate, and the recognition accuracy is improved. The results show that the improvement effect of SIOU is the best, with a 1% increase in each indicator. Therefore, SIOU is more suitable for human flow detection.